连续型隐马尔可夫模型(HMM)参数与语音识别

ON THE PARAMETERS OF HIDDEN MARKOV MODELS(HMM) AND SPEECH RECOGNITION

提出了一种新的连续型隐马尔可夫模型（ＨＭＭ）的概率密度函数，并导出了一系列的参数寻优迭代公式，与常用的概率密度函数相比，它的运算量较小，且不易产生计算时的上溢与下溢问题，把它用于ＨＭＭ语音识别，效果较好．

In speech (or pattern) recognition,the parameters of hidden Markov models play an importante role,and inflence the rate of recognition derectly.In this paper,the new probability density function is defined,and a series iterative formulas are derived.The experiments of speech recognition prove that formulas are more valuable.